Short-term Prediction Of Multiple Loads For Regional Integrated Energy Systems Based On Machine Learning

In the new era,China,as a big energy country,pays more attention to the green and sustainable development of energy,and the regional integrated energy system,as an effective form to promote the construction of China’s energy system,has thus ushered in a new development opportunity.In order to ensure the effective development of system operation management,capacity allocation,optimization and scheduling,it is necessary to accurately predict the energy demand of multiple types of energy in the system.However,compared with the load prediction of single energy in common energy system,the regional integrated energy system is connected with various closely related energy systems,resulting in complex coupling relationship between various energy loads.But the existing load prediction method of single energy is difficult to deal with this complex coupling relationship,which makes the accurate prediction of multiple loads of regional integrated energy system more challenging.Therefore,this paper proposes a short-term multivariate load forecasting method for regional integrated energy systems based on machine learning.The proposed method mainly consists of two parts:one is to determine the input characteristics of the prediction model through quantitative analysis of the coupling relationship between multiple loads;the other is to construct the prediction model reasonably and effectively.First of all,in order to effectively quantify the complex coupling relationship between multiple loads and reasonably select the model input characteristics,this chapter conducts indepth research and analysis on the correlation between RIES ’multiple loads and proposes a method for selecting the input characteristics of a multiple load prediction model.Taking the experimental data used in this paper as an example,the proposed method firstly qualitatively analyzes the main factors that affect multiple loads through the law of load change,then conducts correlation measurement on the above main factors based on AHP and Copula theory,and selects the model input features according to the final correlation measurement results.Then,in order to effectively deal with the complex coupling relationship between multiple types of energy in the regional integrated energy system and realize the objective of accurate prediction of multiple loads,this paper constructs a multiple load prediction model based on GRU and MTL,the proposed model uses the hidden layer parameters of MTL hard Shared mechanism,and the sharing layer is set up by GRU.The proposed model can predict more accurate by realizing the energy complex coupling between the effective learning of information to establish clear between multiple input and multiple output mapping relation.Finally,a case study of a regional integrated energy system is given to verify the effectiveness of the proposed method.At the same time,a comparative experiment is designed to compare and analyze the proposed model with the single-task prediction model and the multioutput prediction model.The experimental results of electric-cold load prediction,electric-heat load prediction,electric-cold load prediction,electric-cold load prediction and electric-cold load prediction show that the proposed model is effective.In addition,combining with NI LabVIEW’s convenient graphical programming function,Python’s powerful rapid development function and MySQL database’s excellent data storage function,this paper design and develop a set of multiple load forecasting system,which mainly realizes the real-time prediction function.The proposed system has a high applicability in practical engineering projects.